LLM Agent Engineer - Specialist - AI Trainer

Obsidian; Miami (FL); On-site; USD 120,000 - 190,000; Full time

Job summary

Obsidian is seeking engineers to build and operate LLM agents in production, with real visibility into how agents are used inside a company.

You have shipped an agent used by real users, and you know how to tell whether changes improved it. You will work on long-running agents, internal monoagents, and reusable skills that connect data and playbooks, while tracking costs and performance across systems.

Qualifications

• Experience shipping an agent that real users relied on.
• Ability to measure agent performance after changes.
• Experience running agents beyond a single request with schedules or long-running tasks.

Responsibilities

• Build and operate LLM agents in production with real visibility into usage.
• Design and discuss how teams evaluate agent reliability and adoption.
• Share concrete war stories about tradeoffs and failures in real systems.

Skills

LLM agent development

Production-grade systems

A/B testing & evaluation

Long-running workflows

Observability & cost tracking

Job description

We are looking for engineers who build and operate LLM agents in production, and who have real visibility into how agents are actually used inside a company.

You have probably:

• Shipped an agent that real users depended on, and been on the hook when it broke.
• Figured out how to tell whether an agent got better or worse after a change.
• Run agents that outlive a single request: scheduled jobs, long-running work, cloud sandboxes.
• Watched your org build an internal assistant, and seen who adopted it and who quietly did not.

We are especially interested in the layers most people do not talk about: internal monoagents wired into company data, shared company memory, reusable skills and playbooks, the tool and MCP surfaces agents call, and how anyone sees what agents did and what they cost.

Applying starts with a short conversational AI interview. No coding, no take-home. We want to hear how you actually think about agent reliability, evaluation, and adoption, and the tradeoffs you have made in real systems. Bring war stories. The messier and more specific, the better.

If that screen stands out, we will invite you to a live 30 minute conversation with our team. We pay $100 to $500 for that conversation, paid on completion of the call, with the amount depending on depth of experience.